3 ms·

This has to be one of the less plausible ways of getting a representative sample that I can think of. Even among the Blizzard employees who happen to be on Bli

This has to be one of the less plausible ways of getting a representative sample that I can think of. Even among the Blizzard employees who happen to be on Blind, most did not answer. So, even if we assume that Blind has a representative sample of Blizzard employees (which seems implausible), the real approval rate could be anywhere from (430+45)/500=95%, or as low as (45/500)=9%. We don't know if those Blizzard employees in favor of the punishment would be more or less likely to respond than others, but it seems unlikely that it would be unrelated. Better claim would be, "Blizzard employees' support for punishment of HK advocate between 9% and 95%". Except, given that Blind's 500 Blizzard employees are probably NOT representative, it might be broader than that.

ladberg 7y ago
Why wouldn't the employees on Blind not be representative? Considering how every other company voted the other way, I don't don't see why Blizzard would be the only biased one.
- CydeWeys 7y ago
  I'm on my company's Blind, and it is vastly different than the company as a whole. It's full of alt-righters, misogynists, xenophobes, transphobes, and other minority viewpoints (within the tech world) that are simply not said out loud at work (or with your name attached online). It drives most reasonable people away to the point that if you tried to use Blind to get an idea of the average viewpoint of employees of the company, you'd be very very off.
  - ninth_ant 7y ago
    But even you are implicitly making the case that your cohort of Blind users is likely the same as in others. To whatever degree that Blind attracts a biased sample within a company, that is true everywhere and even so Blizzards results are a strong outlier. Your excuse would only be plausible if the Blind participation bias only affected Blizzard in some way.
    - CydeWeys 7y ago
      The point I'm making is that Blind is likely not representative of the company at large at all, which is highly relevant here.
diziet 7y ago
I think your math is off - with a 99% confidence interval, on a population of 4,700 employees of Blizzard a sampling of 60 people with a 64.27% approval rate gives us a confidence interval of +-16%~. At 95% confidence, the interval shrinks to 12%~
- JumpCrisscross 7y ago
  > with a 99% confidence interval Confidence intervals assume a random sample. This wasn’t a random sample.
- chongli 7y ago
  You can’t construct that confidence interval unless you know the sample was selected randomly among all Blizzard employees. If the sample is self selected then you don’t have anything at all, really the data is useless. In general, there is a lot of pressure for employees to conform to company decisions, regardless of their actual opinion, because their job is on the line. It does not matter that this “Blind” service purports to be anonymous. People will still demonstrate a bias because they have no guarantee that their communications are not being monitored and ultimately they have nothing to gain by giving their honest opinion.
hbosch 7y ago
I think you may need to revisit your assumption that Blizzard employees are pro-Hong Kong. Why do you think that Blizzard as a company, and Blizzard as a cohort of people, wouldn’t be pro-China?

ladberg 7y ago

Why wouldn't the employees on Blind not be representative? Considering how every other company voted the other way, I don't don't see why Blizzard would be the only biased one.

CydeWeys 7y ago
I'm on my company's Blind, and it is vastly different than the company as a whole. It's full of alt-righters, misogynists, xenophobes, transphobes, and other minority viewpoints (within the tech world) that are simply not said out loud at work (or with your name attached online). It drives most reasonable people away to the point that if you tried to use Blind to get an idea of the average viewpoint of employees of the company, you'd be very very off.
- ninth_ant 7y ago
  But even you are implicitly making the case that your cohort of Blind users is likely the same as in others. To whatever degree that Blind attracts a biased sample within a company, that is true everywhere and even so Blizzards results are a strong outlier. Your excuse would only be plausible if the Blind participation bias only affected Blizzard in some way.
  - CydeWeys 7y ago
    The point I'm making is that Blind is likely not representative of the company at large at all, which is highly relevant here.

diziet 7y ago

I think your math is off - with a 99% confidence interval, on a population of 4,700 employees of Blizzard a sampling of 60 people with a 64.27% approval rate gives us a confidence interval of +-16%~. At 95% confidence, the interval shrinks to 12%~

JumpCrisscross 7y ago
> with a 99% confidence interval Confidence intervals assume a random sample. This wasn’t a random sample.
chongli 7y ago
You can’t construct that confidence interval unless you know the sample was selected randomly among all Blizzard employees. If the sample is self selected then you don’t have anything at all, really the data is useless. In general, there is a lot of pressure for employees to conform to company decisions, regardless of their actual opinion, because their job is on the line. It does not matter that this “Blind” service purports to be anonymous. People will still demonstrate a bias because they have no guarantee that their communications are not being monitored and ultimately they have nothing to gain by giving their honest opinion.

hbosch 7y ago

I think you may need to revisit your assumption that Blizzard employees are pro-Hong Kong. Why do you think that Blizzard as a company, and Blizzard as a cohort of people, wouldn’t be pro-China?